SLIDE 1
DeepGene: An Advanced Cancer Type Classifier Based
- n Deep Learning and Somatic Point Mutations
DeepGene: An Advanced Cancer Type Classifier Based on Deep Learning - - PowerPoint PPT Presentation
DeepGene: An Advanced Cancer Type Classifier Based on Deep Learning - - PowerPoint PPT Presentation
DeepGene: An Advanced Cancer Type Classifier Based on Deep Learning and Somatic Point Mutations (Shi, Yi) 2016.10.03 Center for Systems Biomedicine Shanghai Jiao Tong University Outline Motivation Methods Results & Discussion
SLIDE 2
SLIDE 3
Motivation
SLIDE 4
Motivation Traditional cancer diagnosis
- Morphological appearance,
imaging techniques
- Gene expression
- Protein profiling
Image from radiology.med.nyu.edu Image from well.ox.ac.uk Image from sigmaaldrich.com
SLIDE 5
Motivation Inside drives
- Somatic point mutations
- Insertions and deletions (INDELs)
- Chromatin translocations
- Copy number abnormalities
SLIDE 6
Motivation
Neural network (1940’s) Support vector machine (1960’s) Deep neural network (1980’s)
Supervised combined with un-supervised
SLIDE 7
Motivation
Applications of deep neural network (DNN) learning
SLIDE 8
Methods
SLIDE 9
Methods Three steps of DeepGene
- Step1. Clustered gene filtering (CGF)
- Step2. Indexed sparsity reduction (IDS)
- Step3. Deep neural network (DNN) classifier
SLIDE 10
Methods
- Step1. Clustered gene filtering (CGF)
- Intuitive idea:
Vs.
Team A: Team B:
SLIDE 11
Methods
- Step1. Clustered gene filtering (CGF)
SLIDE 12
Methods
- Step2. Indexed sparsity reduction (ISR)
Raw gene data Nx1
⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎡𝟐 𝟏 𝟐 𝟏 ⋮ 𝟏 𝟐⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎤ * 𝟐 𝟒 ⋮ 𝑶
- ⋮
Indexed gene data nNZx1 if nNZ≥ nISR if nNZ<nISR Truncate the top nISR elements with highest occurrence frequency Add zeros to tail
. 𝟐 𝟒 ⋮ /
⎣ ⎢ ⎢ ⎢ ⎢ ⎡𝟐 𝟒 ⋮ 𝑶 𝟏 ⋮ 𝟏⎦ ⎥ ⎥ ⎥ ⎥ ⎤
Gene data after ISR nISRx1
SLIDE 13
Methods
- Step3. Deep neural network classifier
SLIDE 14
Methods Overall flowchart of DeepGene
Raw gene data N× 1 Clustered gene filtering (CGF) Clustered discriminatory gene data n1×1 Indexed sparsity reducing (ISR) Indexes of non-zero elements n2×1 Concatenation Input to the DNN classifier (n1+n2)×1 DNN classifier
6
Classification label 1×1
KIRP
Classification result (cancer type)
⎣ ⎢ ⎢ ⎢ ⎡𝟐 𝟏 𝟐 𝟏 ⋮⎦ ⎥ ⎥ ⎥ ⎤
⎣ ⎢ ⎢ ⎢ ⎡𝟐 𝟏 𝟏 𝟏 ⋮⎦ ⎥ ⎥ ⎥ ⎤ ⎣ ⎢ ⎢ ⎢ ⎡ 𝟐 𝟒 𝟐𝟏 𝟑𝟗 ⋮ ⎦ ⎥ ⎥ ⎥ ⎤ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎡ 𝟐 𝟏 𝟏 ⋮ 𝟐 𝟒 𝟐𝟏 ⋮ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎤
SLIDE 15
Results & Discussion
SLIDE 16
Results & Discussion Dataset
- 12 tumor somatic point mutation datasets from
- TCGA. (ACC, BLCA, BRCA, CESC, HNSC, KIRP,
LGG, LUAD, PAAD, PRAD, STAD, UCS)
- 22,834 genes from 3,122 samples in total.
Note: ACC, adrenocortical carcinoma; BLCA, bladder urothelial carcinoma; BRCA, breast invasive carcinoma; CESC, cervical squamous cell carcinoma and endocervical adenocarcinoma; HNSC, head and neck squamous cell carcinoma; KIRP , kidney renal papillary cell carcinoma; LGG, brain lower grade glioma; LUAD, lung adenocarcinoma; PAAD, pancreatic adenocarcinoma; PRAD, prostate adenocarcinoma; STAD, stomach adenocarcinoma; UCS, uterine carcinosarcomas.
SLIDE 17
Results & Discussion Parameters
(a) (b) (c) (d)
(a) Parameter estimation for and , corresponding to Table 4; (b) parameter estimation for layer number and parameter number per layer for the DNN classifier, corresponding to Table 5; (c) parameter estimation for cost and gamma for SVM, corresponding to Table 6; (d) parameter estimation for Table 7.
SLIDE 18
Results & Discussion Does CGF and/or ISR help?
10-fold cross validation accuracy of DeepGene with different design options
SLIDE 19
Results & Discussion Comparing to other famous classifiers
Testing accuracy of DeepGene against three widely adopted classifiers
SLIDE 20
Results & Discussion Further investigation
- Integrating other heterogeneous mutation data, e.g.
INDEL, CNV, translocation
- What feature (gene) combinations contribute to
better prediction accuracy? Why?
How this can help real diagnosis?
- Applying to CTC or ctDNA for early diagnosis,
subtyping, locating.
SLIDE 21